Fine-scale population recombination rates, hotspots, and correlates of recombination in the Medicago truncatula genome

Abstract

Recombination rates vary across the genome and in many species show significant relationships with several genomic features, including distance to the centromere, gene density, and GC content. Studies of fine-scale recombination rates have also revealed that in several species, there are recombination hotspots, that is, short regions with recombination rates 10-100 greater than those in surrounding regions. In this study, we analyzed whole-genome resequence data from 26 accessions of the model legume Medicago truncatula to gain insight into the genomic features that are related to high- and low-recombination rates and recombination hotspots at 1 kb scales. We found that high-recombination regions (1-kb windows among those in the highest 5% of the distribution) on all three chromosomes were significantly closer to the centromere, had higher gene density, and lower GC content than low-recombination windows. High-recombination windows are also significantly overrepresented among some gene functional categories-most strongly NB-ARC and LRR genes, both of which are important in plant defense against pathogens. Similar to high-recombination windows, recombination hotspots (1-kb windows with significantly higher recombination than the surrounding region) are significantly nearer to the centromere than nonhotspot windows. By contrast, we detected no difference in gene density or GC content between hotspot and nonhotspot windows. Using linear model wavelet analysis to examine the relationship between recombination and genomic features across multiple spatial scales, we find a significant negative correlation with distance to the centromere across scales up to 512 kb, whereas gene density and GC content show significantly positive and negative correlations, respectively, only up to 64 kb. Correlations between recombination and genomic features, particularly gene density and polymorphism, suggest that they are scale dependent and need to be assessed at scales relevant to the evolution of those features.

FIG. 1.—

Plot of ρ per kilobase along chromosomes 2, 3, and 5 (dark gray) shows variation between the left and right arms (black vertical line is unsequenced centromeric region) as well as distinct variation across chromosomes. Red dots are −log10 of P-values calculated from sequenceLDhot LR scores where LR of 21 is the cutoff for significance. Blue triangles represent NB–ARC genes containing windows with significant hotspot LR scores, and green triangles (Chr3L and Ch5R only) are LRR’s windows with significant LR scores.

FIG. 2.—

Distribution of ρ across chromosomes 2, 3, and 5. Red is 0 median (lower 50% tail), and blue is upper 5% tail.

FIG. 3.—

Mean comparisons of gene density (a), distance to centromere (b), GC content (c), and θ_w (d) between regions of high ρ (upper 5% tail = blue), lower 50% tail (red), and estimated hotspots (green). Error bars represent 95% standard errors.

FIG. 4.—

Smoothed wavelet correlations between recombination (ρ), relative distance to the centromere, gene density, and GC content. Red indicates positive linear relationship and blue indicates negative linear relationship (using t-tests to determine −log₁₀ of P-values). Magnitude of the color is proportional to the level of significance.